Prevent Underfitting and Overfitting for your model

Machine learning and deep learning algorithms are mainly based on statistics. Matching new, unseen data, to the ones we already learned from. This is why there are two typical problems when you train a neural network for a deep learning application. We will take the example of a classifier in this article, but these problems apply to all types of machine learning tasks. The first one is called underfitting, where your model is too simple to represent your data.